The goal for the final project is to evaluate some networking technology issue or proposed new networking standard.Here are some suggested topics, but feel free to pick one on your own.For example, you could expand on any of the issues we touch on in class discussions.Projects can be done individually or in groups of 2 or 3.The project is due on the last day of finals.
1. IP's 32 bit global address space is almost completely allocated, leaving little room for new machines.Two (not necessarily opposing) proposals to deal with this are IPv6 and network address translation.Assuming you were the dictator of the Internet, which technology would you use to deal with the problem, and which do you think the market will pick?
2. Many web services are replicated, both locally across a cluster and geographically across the wide area, in order to provide better availability to end-users.For these services, the Internet is itself the principal availability bottleneck.What should we do to improve the Internet's end to end availability?
3. A malicious attacker has several avenues to causing widespread havoc in the Internet.For example, if the attacker gained control over a router, they could advertise low cost routes to every destination, and then discard all the packets routed through them.Propose a solution!
4. Several commercial TCP "accelerators" are available on the market.Pick one and evaluate it -- how does it work?If it shares a link with a traditional TCP connection, which will get more bandwidth?Is its behavior stable - suppose everyone used the product in place of traditional TCP, would the Internet suffer congestion collapse?
5. A problem with TCP congestion control is that hosts that ignore congestion control signals get better service.This has led some to suggest that gateway routers (e.g., at the campus border) should enforce TCP behavior, by observing packet sends and drops (inferred from acks in the reverse direction), and then making sure that the host sends no faster than TCP would.Demonstrate whether this is practical by trying to implement it.
6. The ATM approach to congestion control is completely different from the Internet's. Assuming you were the dictator of the Internet, which technology would you use to deal with the problem, and which do you think the market will pick?
7. The Internet allows each administrative domain (autonomous system) to choose its own routing policy; the BGP protocol is used to negotiate routes between autonomous systems.Several proposals have been made for supporting multicast across autonomous systems.Which is better?
8. Internet router manufacturers are adding priority scheduling to "support" real-time multimedia traffic.Presumably the real-time traffic will get priority and everyone elese won't.Do you think this will work?
9. The Internet currently uses flat rate pricing, independent of the number of packets sent or the distance a packet must travel to its destination.The telephone network, however, uses flat rate pricing for local calls, but usage-based and distance-based pricing for long-distance calls.Do you think the Internet model is sustainable, particularly when it starts competing with the telephone network by carrying voice calls?
10. Steve McCanne and others at Berkeley have proposed developing a custom transport protocol for web pages, called WebTP.Is this a good or bad idea?
11. Currently, a multicast transmission is by default broadcast to the entire world, and then pruned back if there are no listeners in a given area.One solution to short-circuit pruning is to use the TTL field to scope the multicast transmission to the local area.The TTL field is likewise used for scoping in scalable reliable multicast.Suggest some alternative approaches, and explain which one is best.
12. Explain why topology-aware multicast (or real-time, or caching) algorithms are more efficient.How would you propose the Internet support topology-sensitive applications?
13. A hot area of research is "ad hoc" routing, where instead of having a fixed set of routers, a collection of mobile, wireless hosts can self-organize into a routed fabric. (Think robot swarms on Mars.)How should routing work in this environment?
14. One of the technical problems with ATM was the complexity of creating and tearing down virtual circuits, particularly after some portion of the network failed. The Internet community claimed that this was because ATM virtual circuits were "hard state" that had to be explicitly reclaimed after failures (even though there might not be connectivity to the routers on the far side of the failure).Implement a completely "soft state" version of virtual circuits, where the forwarding tables are reclaimed if not refreshed.How do you efficiently re-route a "soft virtual circuit" around a failure?
Document converted from word 8 by
MSWordView (mswordview 0.5.6)
MSWordView written by Caolan McNamara